Shifting Preference between Oviposition vs. Host-Feeding under Changing Host Densities in Two Aphelinid Parasitoids

Destructive host-feeding is common in hymenopteran parasitoids. Such feeding may be restricted to host stages not preferred for oviposition. However, whether this is a fixed strategy or can vary according to resource levels or parasitoid needs is less clear. We tested the trade-off between host feeding and oviposition on two whitefly parasitoids under varying host densities. Females of two aphelinid parasitoids, Eretmocerus hayati and Encarsia sophia were exposed to nine different densities of their whitefly host, Bemisia tabaci, in single-instar tests to identify their functional response. Mixed-instar host choice tests were also conducted by exposing whiteflies at four densities to the parasitoids. We hypothesized that the parasitoid females can detect different host densities, and decide on oviposition vs. host-feeding accordingly. The results showed that both Er. hayati and En. sophia females tended to increase both oviposition and host-feeding with increased host density within a certain range. Oviposition reached a plateau at lower host density than host-feeding in Er. hayati, while En. sophia reached its oviposition plateau at higher densities. At low densities, Er. hayati parasitized most on first and second (the optimal ones), and fed most on third nymphal instars (the suboptimal one) of the whitefly host as theory predicts, while at high densities, both parasitism and host-feeding occurred on first and second instars which are preferred for oviposition. En. sophia parasitized most on third and fourth (the optimal ones), while fed on first instars (the suboptimal one) at low densities, and utilized third and fourth instars for both at high densities. In conclusion, oviposition vs. host-feeding strategy of parasitoid females was found to vary at different host densities. The balance between reserving optimal hosts for oviposition or using them for host-feeding depended on parasitoid life history and the availability of host resources.


Introduction
Parasitoids have attracted considerable attention because of their importance in biological control of pests, but also due to their value as experimental models in investigating the evolution of reproductive strategies. When confronting a host, the parasitoid females make a series of decisions that have major consequences for the fitness of their offspring [1]. Besides oviposition, many parasitoid species use hosts for food as well. This host-feeding behaviour has been observed in more than 140 species belonging to 17 hymenopteran families [2]. Host feeding provides nutrients for maturing eggs [2][3][4] and prolongs female longevity [3,5] potentially increasing future reproduction.
On the other hand, host-feeding kills the host or at least reduces the quality of the host for oviposition, and the host handling time is often longer than that of ovipositing, increasing the risk of predation to the parasitoid [4]. On encountering a host, the parasitoid should decide whether to oviposit or to host-feed. This decision is affected by factors related to the physiological state of the parasitoid, such as egg load, nutritional status and life expectancy, as well as factors related to host quality and availability [6][7][8][9].
Optimal foraging models predict that host-feeding parasitoid females, when attacking different types of hosts of varying quality, should oviposit on hosts of high quality, while feeding on ones of lower quality [3,10], resulting in more efficient use of host resources [11]. Such behaviour is reported in several species, including Encarsia formosa (Gahan) (Hymenoptera: Aphelinidae) [12], Aphytis lingnanensis (Compere) (Hymenoptera: Aphelinidae) [6], and Diadromus subtilicornis (Gravenhorst) (Hymenoptera: Ichneumonidae) [13]. However, these studies were conducted either in no-choice experiments involving a single host stage, or in mixed-host-stage choice-experiment at constant host densities. The question whether host stage preference for ovposition vs. hostfeeding varies with host density has not yet been explored.
In order to better understand host instar allocation for oviposition vs. host-feeding in response to different host densities, we exposed two aphelinid parasitoids to a variable density host system. Single-instar no choice tests with different host densities were conducted to establish the functional response of both behaviours. Subsequently, mixed-instar host choice tests were conducted to test the host instar allocation ability of the parasitoid females. We hypothesized that the parasitoid females can detect different host densities, and vary their oviposition vs. host-feeding accordingly. Under abundance of high quality hosts (at high host density), a parasitoid female is expected to both oviposit and hostfeed on its favourite host stages. When host resource is limited (at low host density and/or quality), a parasitoid female would refrain from host feeding on the optimal host instar for oviposition, and feed on alternative host instars.
Er. hayati is a solitary parasitoid ovipositing externally under whitefly nymphs [20]. Upon eclosion, the first instar larva penetrates the host cuticle, feeds and pupates internally. This species is a biparental primary parasitoid, with both males and females developing in whitefly nymphs.
Both En. sophia and Er. hayati females oviposit and host-feed on all nymphal instars of B. tabaci with the exception of late fourth instars. Third and fourth instars are the optimal hosts for En. sophia oviposition, while Er. hayati oviposits mostly on first and second instars [15,21].

Experimental parasitoids
B. tabaci-infested tomato plants were exposed to naïve female wasps (up to 1-day old) for 24 h. Subsequently, plants were maintained in an air-conditioned laboratory, where all experiments were conducted, at 2661uC, 6565% RH and a light regime of 14:10 (hours L:D). After 13 days, parasitoid pupae were collected and individually put in a Petri dish with a drop of honey (5%). Petri dishes were checked daily for emergence. Females that emerged were provided with males and were observed mating, then used in pair in experiments 1 day later. All experimental females had no ovipositing experience.

Single-instar no-choice tests
In these experiments, only the single (optimal) nymphal instar of B. tabaci was offered to the parasitoids: second instar nymphs (N2) to Er. hayati, and third instar nymphs (N3) to En. sophia females. In order to generate the required nymphal population, 10, 20, 30 or 40 unsexed whitefly adults were introduced into a clip cage (transparent plastic cup, base diameter 2.1 cm) on a tomato leaf (one clip cage per plant) for a 12-h oviposition period to assure host stage uniformity. The adults were removed and the eggs were monitored daily until they developed to the desired stage (N2 or N3). Five, 10, 20, 30, 40, 50, 60, 70 or 80 nymphs were left on each leaf in the area covered by a clip cage (3.5 cm 2 ). A pair of 1 d-old Er. hayati or En. sophia was introduced into a clip cage confining the nymphs in each of the nine host densities for a 24-h oviposition and feeding period, after which they were removed. The numbers of whitefly hosts that were killed by parasitizing or host-feeding were checked 7-8 days after parasitoid removal. If the hosts were parasitized, the mycetome displacement was visible through the cuticle at the time of examination [22]. If the hosts were killed by host-feeding, the bodies became flat, desiccated and colour faded [23]. The hosts killed by attempting (but failed) hostfeeding or dead naturally, identified by the appearance of bodies that contain inclusions and turn darker because of no/less body fluid is lost, were excluded from the count. The experiment had a total of 18 treatments (nine host densities and two parasitoid species). A total of 15 pairs for each treatment were initially used. The replicates in which the introduced parasitoids escaped or died were excluded, and the data for at least10 replicates were used in data analysis.

Mixed-instar host choice tests
Ten to 20 unsexed whitefly adults were introduced into a clip cage on a tomato leaf (one clip cage per plant) for a 12-h oviposition period, then removed to assure stage uniformity. The whitefly adults were introduced on the same tomato leaf for 1, 3 and 6 days after the first removal of the introduced adults, respectively. The whitefly eggs were monitored daily until they developed to the desired stage, namely first (N1), second (N2), third (N3) and early fourth instars (N4). The four nymphal instars of B. tabaci were offered to the parasitoids simultaneously. A total of 20, 40, 60 and 80 nymphs with different instars in equal proportions were left on each leaf in the area and covered by a clip cage (3.5 cm 2 ). The distributions of the nymphs on each experimental plant leaf were photographed under a stereomicroscope and the photos were printed out. A pair of 1 d-old Er. hayati or En. sophia adults was introduced into a clip cage confining the nymphs in each of the four host densities for a 24-h oviposition and feeding period, then removed. The numbers of whitefly hosts that were killed by parasitizing and host-feeding were checked 7-8 days after parasitoid removal. The host instars killed by parasitizing or feeding were determined by using the photos taken before parasitoid introduction. The experiment had a total of 8 treatments (four host densities and two parasitoid species). A total of 15 pairs for each treatment were initially used. The replicates in which the introduced parasitoids escaped or died were abandoned. The data for at least10 replicates were used in data analysis.

Statistical analysis
In single-instar no choice experiment, the differences in the number of whitefly hosts parasitized, as well as the number of whitefly hosts fed by both species of parasitoids were analyzed using one-way ANOVA (SPSS version 13.0 software package) with the factor of host density. Differences among means related to different host densities were compared with Tukey's Honestly Significant Difference (HSD) test at the significance level p = 0.05.
In mixed-instar choice experiment, a two-way ANOVA was used, with whitefly host instar and host density as factors, to analyze their effects on mean number of hosts killed (the sum of hosts parasitized or host-fed), parasitized, as well as host-fed. If the factor showed significant influence, the differences in the number of whitefly hosts killed, parasitized, or host-fed at each host density were analyzed using one-way ANOVA with the factor of host instar. Differences in the number of whitefly hosts killed, parasitized, or host-fed on each host instar were also analyzed by one-way ANOVA with the factor of host density. Differences among means related to different host instars or different host densities were compared with HSD test at p = 0.05.
The host-feeding ratio was defined as: Host-feeding ratio = no. of hosts fed upon/(no. of host parasitized + no. of host fed upon).
In both single-instar and mixed-instar experiments, differences among means of host-feeding ratio of both species of parasitoids related to the factor of different host densities were compared with Kruskal-Wallis test at p = 0.05. In mixed-instar experiment, differences among means of host-feeding ratio of both species of parasitoids on different host instars were also compared with Kruskal-Wallis test at p = 0.05 with the factor of host instar.
Differences between means of host-feeding ratio by the factor of different species of parasitoids at each host density in single-instar experiment, as well as the differences between means of hostfeeding ratio in single-instar experiment and means of host feeding ratio on each instars in total in mixed-instar experiment with the factor of no-choice or choice condition were compared with Mann-Whitney test at p = 0.05. We used non-parametric methods because host-feeding ratio did not meet the assumption of normal distribution and equal variances for parametric method even after transformation.

Single-instar no choice tests
The number of hosts parasitized by Er. hayati significantly increased with increased density of B. tabaci nymphs (F 8, 81 = 103.70, P,0.0001) and reached a plateau at 11.2 (SE = 0.5) at 40 2 nd instar nymphs per cage ( Figure 1A). The number of hosts fed by Er. hayati also significantly increased with increased host density (F 8, 81 = 72.05, P,0.0001) and reached a plateau at 6.9 (SE = 0.2) at 60 nymphs per cage ( Figure 1A).

Mixed-instar host choice tests
In mixed-instar host test, number of hosts parasitized, number of hosts fed, and the total number of hosts killed by parasitism and host-feeding were significantly affected by host instars and host densities (Table 1). There were significant interactions between host instars and host densities on parasitism and host-feeding of the two parasitoids (Table 1).
Hosts parasitized and fed by Eretmocerus hayati. In mixed-instar host test, the total number of whitefly hosts killed by parasitism and host-feeding by Er. hayati females varied significantly with host instars at all of the four host densities (Figure 2A, F 3, 36 = 13.52, 20.27, 59.03 and 48.41 at 20, 40, 60 and 80 nymphs per cage, respectively, P,0.0001 for each). At 20 nymphs per cage, females killed a similar number of first, second and third host instar nymphs, and significantly fewer fourth instar ones. With the host density increasing to 40, 60 and 80 nymphs per cage, the first (F 3, 36 = 30.92, P,0.0001) and second (F 3, 36 = 11.10, P,0.0001) instar nymphs killed by females increased significantly, while the number of third (F 3, 36 = 1.24, P = 0.308) and fourth (F 3, 36 = 1.09, P = 0.365) instar nymphs changed less (Figure 2A). At 60 and 80 nymphs per cage, female killed most on first instars, followed by second, third and fourth instars, in this sequence (Figure 2A).
The numbers of hosts parasitized by Er. hayati females varied significantly with host instar at all of the four host densities ( Figure 2B, F 3, 36 = 15.31, 21.85, 34.22 and 45.89, at 20, 40, 60 and 80 nymphs per cage, respectively, P,0.0001 for each). At all experimental host densities, Er. hayati parasitized significantly more first and second instars than third and fourth instars, with the exception that the number of third instars parasitized was not significantly less than that of second instars at 40 nymphs per cage ( Figure 2B). The numbers of parasitized first (F 3, 36 = 11.02, P,0.0001) and second instar nymphs (F 3, 36 = 7.09, P = 0.0007) increased significantly with increased host density, while the numbers of parasitized third (F 3, 36 = 1.49, P = 0.234) and fourth instar nymphs (F 3, 36 = 0.28, P = 0.838) did not ( Figure 2B).
At the host density of 20 nymphs per cage, the number of different host instars fed by Er. hayati females also varied significantly (F 3, 36 = 7.89, P = 0.0004): females fed significantly more number of third instars than first and fourth instars, and more second instars than first instars; there was no difference between the number of second and third instars fed by females ( Figure 2C). At 40 nymphs per cage, there were no significant differences among the number of the four different instars fed ( Figure 2C, F 3, 36 = 2.17, P = 0.109), although the number of second and third instars fed was higher than the others. With the host density increased, the number of first (F 3, 36 = 18.98, P,0.0001) and second instar nymphs (F 3, 36 = 4.49, P = 0.009) fed by females increased significantly, while the number of third (F 3, 36 = 0.45, P = 0.718) and fourth instar nymphs (F 3, 36 = 3.144, P = 0.037) did not ( Figure 2C). At 80 nymphs per cage, no fourth instar nymphs were fed upon ( Figure 2C). At 60 (F 3, 36 = 13.04, P,0.0001) and 80 (F 2, 27 = 6.636, P = 0.005) nymphs per cage, Er. hayati female fed most on first and second instar nymphs ( Figure 2C).
Hosts parasitized and fed by Encarsia Sophia. In mixedinstar host test, at 20 nymphs per cage, a similar number of the four host instars were killed ( Figure 3A, F 3, 36 = 2.16, P = 0.109). With increased host density, the number of the third and fourth instar nymphs killed by females increased (F 3, 36 = 29.22 and 28.82, respectively, both P,0.0001), while that of the first instar decreased ( Figure 3A, F 3, 36 = 12.48, P,0.0001). There was little change in the numbers of second instar nymphs attacked ( Figure 3A, F 3, 36 = 1.94, P = 0.140). At 60 and 80 nymphs per cage, female killed most on third and fourth instars, followed by second instar, and least on first instar ( Figure 3A, F 3, 36 = 49.61 and 66.89 at 60 and 80 nymphs per cage, respectively, both P,0.0001).
Numbers of different instar nymphs parasitized by En. sophia females varied significantly at the four host densities ( Figure 3B, F 3, 36 = 19.75, 7.04, 29.03 and 43.59, at 20, 40, 60 and 80 nymphs per cage, respectively, P,0.0008 for each). The females parasitized most on third and fourth instars, followed by second, and least on first intars at 20, 60 and 80 nymphs per cage, while parasitized similar number of third, fourth and second insta r, and less number of first instar at 40 nymphs per cage ( Figure 3B). The numbers of parasitized third and fourth instar nymphs increased significantly (F 3, 36 = 11.59 and 14.86, respectively, both P,0.0001) with increased host density, while the numbers of parasitized first (F 3, 36 = 2.20, P = 0.104) and second (F 3, 36 = 1.95, P = 0.140) instar nymphs did not differ ( Figure 3B). As the host density increased from 20 to 40, the number of each instar parasitized were not significantly increased ( Figure 3B), however, the total increase in parasitism (1.4 nymphs) was mainly due to the increased number of second instars parasitized (0.7 nymphs).
At the host density of 20 nymphs per cage, En. sophia females fed most on the first instar nymphs, followed by second, third and fourth instars with similar numbers ( Figure 3C, F 3, 36 = 5.88, P = 0.002). At 40 nymphs per cage, the number of third instar nymphs fed by females increased significantly from the lower host density ( Figure 3C, F 3, 36 = 16.52, P,0.0001). The total increase in host feeding (1.4 nymphs) was mainly due to the increased number of third (1.0 nymphs) and fourth (0.3 nymphs) instars. With the host density increased to 60 and 80 nymphs per cage, females increasingly allocated their host feeding from first instar to third and fourth instars ( Figure 3C, F 3, 36 = 10.30, 16.52 and 9.69 for N1, N3 and N4, respectively, P,0.0001 for each). At 60 nymphs per cage, En. sophia female fed most on third instar and fourth instars ( Figure 3C, F 3, 36 = 11.21, P,0.0001). At 80 nymphs per cage, En. sophia female fed most on third instar nymphs ( Figure 3C, F 3, 36 = 14.90, P,0.0001).

Discussion
Both Er. hayati and En. sophia females exhibited a tendency to increase both oviposition and host-feeding with increased host density within a certain range. This was observed in other species of parasitoids [24][25][26]. However, the host handling strategy related to varied host densities of Er. hayati and En. sophia showed differences. In single-instar tests, the number of hosts parasitized by Er. hayati increased faster than the number of hosts fed, and reached the maximal number at lower host densities than hostfeeding. For En. sophia, the two behaviours changed similarly to increased host densities, but the number of hosts fed reached a plateau at lower host densities than the number of hosts parasitized. In the present mixed-instar tests, when host density initially increased, most of the increase in parasitism by En. sophia occurred on second instar nymphs, while more than 70% of the increase in host-feeding occurred on third instar which was the optimal instar either for parasitizing or host feeding. In the case of Er. hayati, under increased host density, the majority of increase in both parasitism and host-feeding occurred on the first instar, even if the number of first instars fed upon was not the highest. The differences between these two parasitoids may be explained by the egg load of the females. To get nutrients to make more mature eggs is one of the reasons for host feeding [2][3][4]. The number of mature ova of 1-d-old Er. hayati was higher than that of En. sophia (Yang NW, unpublished data). Consequently, more eggs could be laid by Er. hayati with less host-feeding than En. sophia, and the priority for host-feeding in En. sophia seems reasonable.
In the present study, the proportion of each host instars killed by these two parasitoids varied with different host densities. At low host density, the host encounter rate is also low [4], which possibly forces the parasitoid to compromise and use not only the favorite, but also other host instars. However, the host-feeding ratio on alternative instars was higher than that on optimal instars, suggesting that parasitoid females partition their feeding and oviposition behaviour on different instars as expected on theoretical grounds, and the optimal instars for oviposition were fed upon less [3,[10][11][12][13]. At high host density, the host encounter rate is also high [4], providing opportunity to exercise instar preferences without compromise. Under those conditions, more optimal instars than alternative ones were killed, but the hostfeeding ratio on each instar was similar. This indicates that there was no partition between feeding and oviposition towards different instars. Host-feeding ratio of Er. hayati in the single-instar condition was a little lower than that under mixed-instar presence, and the availability of mixed instars had a smaller effect on host-feeding ratio than that of host density.
Since the host handling strategy varied with host density, as well as by species, the assessment of the preferred instar for oviposition vs. host-feeding needs to be made with caution. When host resources included a mixture of different instars and were abundant, both parasitoids oviposited, but also host-fed mostly on the optimal instars, which obey to their true preference. In the case of limited host resources, host-feeding was switched onto alternative instars, but in a species-specific manner. Thus, under a mixed-instar host availability scenario, the oviposition response is conservative, but the preferred instars for host-feeding could not simply be identified by the number of each instar fed upon since the shortage of suitable instars at low density drives the females to shift the true preference. One has to make sure that the host density is high enough not to compel the females to switch hostfeeding on alternative instars -under limited host availability, these will be the suboptimal ones, masking their true preference. This is not yet obvious in the literature, because most studies were done under a single (even though usually high) host density [8,23,27]. Host-feeding behaviour observed under single-instar host availability may also differ from the more natural, mixedinstar host presence. En. sophia, En. formosa and Eretmocerus melanoscutus (Zolnerowich & Rose) (Hymenoptera: Aphelinidae) host-feed most on first and second instar nymphs of B.tabaci Middle East-Asia Minor 1 in single-instar experiments, while third and fourth instar nymphs are fed upon most when a mixture of different instars are available [23]. Since the co-occurrence of different-instar nymphs on one leaf is the common condition in the field, host-feeding results obtained using a single instar may not well predict the real biocontrol efficiency by host-feeding in the field.
Our findings have consequences for the practice of biological control, especially for augmentative releases. It is very important to find ways to manipulate the parasitoids so that they quickly destroy as many hosts as possible through destructive host feeding and parasitism. The present study suggests that the combined release of these two parasitoids might have better control on B. tabaci than h. An area of 3.5 cm 2 of a leaf on a potted tomato plant was covered by a clip cage. Host instars, N1, N2, N3 and N4 are first, second, third and early fourth instars, respectively. Bar heads with different lowercase letters in each cluster indicate significant differences (HSD test; P,0.05) in number of hosts parasitized or fed among different host instars. Bar heads with different capital letters for each instars between the clusters indicate significant differences (HSD test; P,0.05) in number of hosts killed among different host densities. Vertical bars indicate 6 one SE. doi:10.1371/journal.pone.0041189.g003 that of either species alone, since these two parasitoids not only preferred different host instars to oviposit and host-feed at high host density but also reacted differently to increased host densities. Releasing one species might be efficient at low host densities since oviposition and host-feeding under those conditions would be partitioned on different host instars and no instar will be free of parasitoid pressure. However, when host density is high, attack will shift onto the favorite host instar/s. In such cases, the release of multiple parasitoids which complement each other via differences in preferred instars might be more effective than repeated single parasitoid releases. However, as En. sophia is an autoparasitoid, the interference to Er. hayati population by the male producing hyperparasitizing behavior when released together need to be concerned.
Optimal foraging models predict that parasitoid females oviposit on hosts of high quality, while feed on the ones of lower quality [3,10]. Our study demonstrates that the preference and intensity of oviposition and host-feeding varied with host density. Organisms are under environmental constraints that limit their possibilities to maximize their fitness [28]. Our findings suggest that once the food resource stress is relaxed, the parasitoid strategy of allocating host recourses increases the fitness not only via reproduction, but also via body maintenance. However, the difference in the reaction of females released in patches with different host densities may be not only numerical, further behavioral observation need to be conducted to reveal the mechanism behind this. Proportion of hosts fed upon in 24h by 1 d-old females of Eretmocerus hayati or Encarsia sophia, respectively. An area of 3.5 cm 2 of a leaf on a potted tomato plant was covered by a clip cage. Host instars, N1, N2, N3 and N4 are first, second, third and early fourth instars, respectively. Single-instar tests were performed on N2 for Er. hayati, and N3 for En. sophia. doi:10.1371/journal.pone.0041189.g004